Given the unique cognitive advantages of humans in complex maritime missions, this study proposes an intelligent collaborative decision-making framework for a hybrid system consisting of one crewed surface vehicle and multiple uncrewed surface vehicles (MSV-USV). We introduce a regularized polygonal obstacle modeling method based on geometric feature simplification to reduce computational complexity. A distributed task negotiation framework with dynamic priority-based message-passing enables adaptive task decomposition through multi-agent interactions. For maneuver optimization, a tripartite game model captures multidirectional strategy interactions while ensuring MSV safety requirements. Monte Carlo simulations validate the algorithm’s stability and reliability, and prototype tests demonstrate engineering feasibility. Note to Practitioners—With the increasing application of uncrewed systems in maritime missions, effectively integrating the cognitive advantages of human operators with the computational capabilities of uncrewed systems has become a key challenge in enhancing mission performance. The proposed distributed collaborative decision-making framework is specifically designed for mixed fleets consisting of crewed surface vessels and multiple uncrewed surface vessels, and can be applied to complex and dynamic mission scenarios such as collaborative defense and maritime search and rescue. The framework ensures real-time decision-making through geometrically simplified obstacle modeling and dynamic priority-based distributed task negotiation, while a tripartite game optimization guarantees the superiority. Monte Carlo simulations and prototype tests have demonstrated the framework’s excellent stability, reliability, and engineering applicability. However, the study is based on ideal sensor assumptions and does not yet account for complex noise interference in real-world maritime environments. Future work may extend the framework to cross-domain mixed systems involving underwater, surface, and aerial platforms, while incorporating more robust perception algorithms to improve decision-making reliability in noisy real-world conditions.
